Programming Languages CSCE 314 Programming Languages A Tour of Language Implementation Dr. Hyunyoung Lee

Programming Language Characteristics Different approaches to describe computations, to instruct computing devices E.g., Imperative, declarative, functional Different approaches to communicate ideas between humans E.g., Procedural, object-oriented, domain-specific languages Programming languages need to have a specification: meaning (semantics) of all sentences (programs) of the language should be unambiguously specified

Programming Language Expressiveness Different levels of abstraction More abstract Haskell, Prolog sum[1..100] Scheme, Java mynum.add(5) C i++; Assembly language iadd Machine language 10111001010110

Evolution of Languages 1940’s: connecting wires to represent 0’s and 1’s 1950’s: assemblers, FORTRAN, COBOL, LISP 1960’s: ALGOL, BCPL (→ B → C), SIMULA 1970’s: Prolog, FP, ML, Miranda 1980’s: Eiffel, C++ 1990’s: Haskell, Java, Python 2000’s: D, C#, Spec#, F#, X10, Fortress, Scala, Ruby, . . . 2010’s: Agda, Coq . . . Evolution has been and is toward higher level of abstraction

Defining a Programming Language Syntax: Defines the set of valid programs Usually defined with the help of grammars and other conditions if-statement ::= if cond-expr then stmt else stmt | if cond-expr then stmt cond-expr ::= . . . stmt ::= . . . Semantics: Defines the meaning of programs Defined, e.g., as the effect of individual language constructs to the values of program variables if cond then true-part else false-part If cond evaluates to true, the meaning is that of true-part; if cond evaluates to false, the meaning is that of false-part.

Implementing a Programming Language Task is to undo abstraction. From the source: int i; i = 2; i = i + 7; to assembly (this is actually Java bytecode): iconst_2 // Put integer 2 on stack istore_1 // Store the top stack value at location 1 iload_1 // Put the value at location 1 on stack bipush 7 // Put the value 7 on the stack iadd // Add two top stack values together istore_1 // The sum, on top of stack, stored at location 1 to machine language: 00101001010110 01001010100101

Implementing a Programming Language – How to Undo the Abstraction Source program Optimizer I/O Type checker Code generator Machine code Lexer Parser Machine Bytecode JIT Interpreter Virtual machine I/O I/O

Compiling and Interpreting (1) Typically compiled languages: C, C++, Eiffel, FORTRAN Java, C# (compiled to bytecode) Typically interpreted languages: Python, Perl, Prolog, LISP Both compiled and interpreted: Haskell, ML, Scheme

Compiling and Interpreting (2) Borderline between interpretation and compilation not clear (not that important either) Same goes with machine code vs. byte code. Examples of modern compiling/interpreting/executing scenarios: C and C++ can be compiled to LLVM bytecode Java compiled to bytecode, bytecode interpreted by JVM, unless it is first JITted to native code, which can then be run on a virtual machine such as VMWare

Lexical Analysis From a stream of characters if (a == b) return; to a stream of tokens keyword[‘if‘] symbol[‘(‘] identifier[‘a‘] symbol[‘==‘] identifier[‘b‘] symbol[‘)‘] keyword[‘return‘] symbol[‘;‘]

Syntactic Analysis (Parsing) From a stream of characters if (a == b) return; to a stream of tokens keyword[‘if‘] symbol[‘(‘] identifier[‘a‘] symbol[‘==‘] identifier[‘b‘] symbol[‘)‘] keyword[‘return‘] symbol[‘;‘] to a syntax tree (parse tree) if-statement expression statement equality operator return stmt identifier identifier a b

Type Checking if (a == b) return; Annotate syntax tree with types, check that types are used correctly if-statement : OK statement : OK identifier : int b a expression : bool equality operator : integer equality return stmt : void

Optimization int a = 10; int b = 20 – a; if (a == b) return; Constant propagation can deduce that always a==b, allowing the optimizer to transform the tree: if-statement : OK statement : OK identifier : int b a expression : bool equality operator : integer equality return stmt : void if-statement : OK statement : OK true constant : bool return stmt : void return stmt : void

Code Generation Code generation is essentially undoing abstractions, until code is executable by some target machine: Control structures become jumps and conditional jumps to labels (essentially goto statements) Variables become memory locations Variable names become addresses to memory locations Abstract data types etc. disappear. What is left is data types directly supported by the machine such as integers, bytes, floating point numbers, etc. Expressions become loads of memory locations to registers, register operations, and stores back to memory

Phases of Compilation/Execution Characterized by Errors Detected Lexical analysis: 5abc a === b Syntactic analysis: if + then; int f(int a]; Type checking: void f(); int a; a + f(); Execution time: int a[100]; a[101] = 5;